This specification relates to detecting objects in images.
Deep neural networks are machine learning systems that employ multiple layers of models, where the outputs of lower level layers are used to construct the outputs of higher level layers.